[c-nsp] IPv6 BGP on 7200

[Marcus Stoegbauer]

On Tuesday 27 February 2007 13:11:58 Stefan M. Brandl wrote:
> On Tue, Feb 27, 2007 at 11:10:53AM +0000, Paul Cairney wrote:
> > On Tue, Feb 27, 2007 at 09:40:38AM +0100, Stefan M. Brandl wrote:
> > > I have two Cisco 7206-NPE-G1, connected via OC3.
> > > Both routers do BGP with IPv4 and v6.
> > > The v6-session goes down every 10-30 minutes:
> > >
> > > Feb 27 08:08:18: %BGP-5-ADJCHANGE: neighbor
> > > 2001:16E0:0:5:0:DEAD:BEEF:1 Up Feb 27 08:15:54: %BGP-5-ADJCHANGE:
> > > neighbor 2001:16E0:0:5:0:DEAD:BEEF:1 Down Peer closed the session
> >
> > > The IPv4-BGP-session is stable.
> > > Both routers also do v4 and v6 BGP to other routers just fine.
> > > Where could be the problem? How to debug?
> >
> > Did this occur following an upgrade to 12.2(25)S12 by any chance?
>
> Yes, one router is running 12.2(25)S12

Sorry for the late reply, but we ran into the same issue with 12.2(25)S12 
(started with S11) too. Cisco TAC tracks this as CSCsh82879 and marked it 
as a duplicate of CSCsc39357.
Unfortunately the workaround for CSCsc39357 (increasing the tcp window size) 
only increases the time between the BGP flaps, but doesn't make them go 
away. 
Interesting fact: Those flaps also affect v4 BGP sessions, but only on a 
wider scale, for example if you change MED on many prefixes at once.

On the positive side, the fix for this issue shall be built into 12.2
(25)S13, which is due to release in April (this year, I hope).

   Marcus